Merge branch 'master' into asyncWaiterTimeoutHandle

Author: nargokul

CHANGELOG.md

@@ -1,5 +1,20 @@
 # Changelog
 
+## v2.237.3 (2025-01-09)
+
+### Bug Fixes and Other Changes
+
+ * pin metadata-version to 2.3
+ * model server might have already done a serialization. honor that by not decoding the request again if it is not already bytes or bytestream
+ * Disable jumpstart tests missing clean up logic
+ * Jumpstart ap southeast 5
+ * add autogluon 1.2
+ * updated inference script to cover context
+ * security update -> use sha256 instead of md5 for file hashing
+ * Fix Flake8 Violations
+ * Added parsing string support for situations where custom code might be used (ie. mlflow)
+ * Updating Inference Optimization Validations
+
 ## v2.237.2 (2024-12-17)
 
 ### Bug Fixes and Other Changes

VERSION

@@ -1 +1 @@
-2.237.3.dev0
+2.237.4.dev0

pyproject.toml

@@ -73,13 +73,15 @@ pattern = "(?P<version>.+)"
 [tool.hatch.metadata.hooks.custom]
 
 [tool.hatch.build.targets.wheel]
+core-metadata-version = "2.3"
 packages = ["src/sagemaker"]
 exclude = ["src/sagemaker/serve/model_server/triton/pack_conda_env.sh"]
 
 [tool.hatch.build.targets.wheel.shared-scripts]
 "src/sagemaker/serve/model_server/triton/pack_conda_env.sh" = "pack_conda_env.sh"
 
 [tool.hatch.build.targets.sdist]
+core-metadata-version = "2.3"
 only-include = [
   "/requirements/extras",
   "/src",

src/sagemaker/modules/local_core/local_container.py

@@ -108,6 +108,8 @@ class _LocalContainer(BaseModel):
     container_entrypoint: Optional[List[str]]
     container_arguments: Optional[List[str]]
 
+    _temporary_folders: List[str] = []
+
     def model_post_init(self, __context: Any):
         """Post init method to perform custom validation and set default values."""
         self.hosts = [f"algo-{i}" for i in range(1, self.instance_count + 1)]
@@ -201,6 +203,13 @@ def train(
 
         # Print our Job Complete line
         logger.info("Local training job completed, output artifacts saved to %s", artifacts)
+
+        shutil.rmtree(os.path.join(self.container_root, "input"))
+        shutil.rmtree(os.path.join(self.container_root, "shared"))
+        for host in self.hosts:
+            shutil.rmtree(os.path.join(self.container_root, host))
+        for folder in self._temporary_folders:
+            shutil.rmtree(os.path.join(self.container_root, folder))
         return artifacts
 
     def retrieve_artifacts(
@@ -540,6 +549,7 @@ def _get_data_source_local_path(self, data_source: DataSource):
             uri = data_source.s3_data_source.s3_uri
             parsed_uri = urlparse(uri)
             local_dir = TemporaryDirectory(prefix=os.path.join(self.container_root + "/")).name
+            self._temporary_folders.append(local_dir)
             download_folder(parsed_uri.netloc, parsed_uri.path, local_dir, self.sagemaker_session)
             return local_dir
         else:

tests/integ/sagemaker/modules/train/test_local_model_trainer.py

@@ -92,10 +92,7 @@ def test_single_container_local_mode_local_data(modules_sagemaker_session):
                 "compressed_artifacts",
                 "artifacts",
                 "model",
-                "shared",
-                "input",
                 "output",
-                "algo-1",
             ]
 
             for directory in directories:
@@ -149,14 +146,16 @@ def test_single_container_local_mode_s3_data(modules_sagemaker_session):
             assert os.path.exists(os.path.join(CWD, "compressed_artifacts/model.tar.gz"))
         finally:
             subprocess.run(["docker", "compose", "down", "-v"])
+
+            assert not os.path.exists(os.path.join(CWD, "shared"))
+            assert not os.path.exists(os.path.join(CWD, "input"))
+            assert not os.path.exists(os.path.join(CWD, "algo-1"))
+
             directories = [
                 "compressed_artifacts",
                 "artifacts",
                 "model",
-                "shared",
-                "input",
                 "output",
-                "algo-1",
             ]
 
             for directory in directories:
@@ -204,20 +203,20 @@ def test_multi_container_local_mode(modules_sagemaker_session):
 
             model_trainer.train()
             assert os.path.exists(os.path.join(CWD, "compressed_artifacts/model.tar.gz"))
-            assert os.path.exists(os.path.join(CWD, "algo-1"))
-            assert os.path.exists(os.path.join(CWD, "algo-2"))
 
         finally:
             subprocess.run(["docker", "compose", "down", "-v"])
+
+            assert not os.path.exists(os.path.join(CWD, "shared"))
+            assert not os.path.exists(os.path.join(CWD, "input"))
+            assert not os.path.exists(os.path.join(CWD, "algo-1"))
+            assert not os.path.exists(os.path.join(CWD, "algo-2"))
+
             directories = [
                 "compressed_artifacts",
                 "artifacts",
                 "model",
-                "shared",
-                "input",
                 "output",
-                "algo-1",
-                "algo-2",
             ]
 
             for directory in directories: